1. Relational and procedural accounts

To understand what a relational account of intellectual autonomy might look like, it will help to be more precise about the nature of relational views of practical autonomy. In the broadest sense, relational views of autonomy are those that acknowledge social relations that are either:

1. (a) compatible with individual autonomous agency,

2. (b) causally salient for developing or maintaining autonomous agency, or

3. (c) conceptually necessary for being an autonomous agent.

Since (a) expresses a rejection of individualistic views which equate autonomy with self-reliance, I include it in the broadest sense of the term.Footnote ² Usually, however, ‘relational autonomy’ refers to views committed to either (b) or (c), and I will use the term in this narrower sense moving forward. Commitment to (b) is sufficient for a specifically causally relational view of autonomy, and commitment to (c) for a constitutively relational view of autonomy (Mackenzie and Stoljar Reference Mackenzie, Stoljar, Mackenzie and Stoljar2000, 22).

Although they are not mutually exclusive, relational views are typically developed in response to purely procedural views of autonomy. On procedural accounts of autonomy, the contents of an individual’s psychological attitudes are irrelevant to whether the choices or actions that follow from them are autonomous. What matters is the agent’s motivational structure or the processes through which their choices were formed. For example, Harry Frankfurt argues that identification with one’s first-order desires in the form of second-order volitions is enough to secure autonomy of the will (Frankfurt Reference Frankfurt1971). The contents of one’s will do not matter for autonomous agency. What matters is whether one has the will one wants to have. In contrast to Frankfurt’s hierarchical account of autonomy, John Christman’s historical account emphasizes critical reflection and revision of one’s desires in order to distinguish autonomy conferring from autonomy undermining processes (Christman Reference Christman1991). What matters for a historical view is not the content of one’s will or the current structure of one’s will but features of the historical process through which it was formed. While these and other procedural accounts all differ in the details, they all in some way or another emphasize capacities for critical reflection on and revision of one’s values or desires (Westlund Reference Westlund2009, 26; Mackenzie and Stoljar Reference Mackenzie, Stoljar, Mackenzie and Stoljar2000, 13–14). Since it’s perfectly possible to rely on others while engaging in the process of critical reflection, procedural accounts are at least compatible with some forms of social dependence, though which sorts of relations are compatible will depend on the specifics of the view.Footnote ³

Causally relational accounts explicitly treat some forms of dependence as not just compatible with but also causally salient for developing and maintaining autonomous agency. Historical examples include the views of autonomy developed in an educational context by John Dewey (Reference Dewey1916) and Paulo Freire (Reference Freire2000) and, arguably, the views of self-determination developed in a post-Kantian context by Johann Fichte (Reference Fichte2000) and G.W.F. Hegel (Reference Hegel1977) (though the latter two are constitutively relational as well). More recently, Jennifer Nedelsky points out that ‘if we ask ourselves what actually enables people to be autonomous, the answer is not isolation, but relationships – with parents, teachers, friends, loved ones’ (Nedelsky Reference Nedelsky1989, 12). Focusing on the negative contrast, Diana Meyers and Marilyn Friedman explore the ways in which oppressive forms of gender socialization can undermine the development of capacities that make autonomous choice and action possible (Meyers Reference Meyers1987, Reference Meyers, Mackenzie and Stoljar2000, Reference Meyers2002; Friedman Reference Friedman2003). In general, causal theorists are primarily interested in the ways adverse social conditions can impair or impede individual autonomy and identifying the conditions that allow autonomy to flourish. They are not, however, interested in including relational conditions in their analyses of autonomy. While this doesn’t mean that they fail to make an important or novel contribution, it does mean that only constitutively relational views offer a genuine alternative to procedural views (Christman Reference Christman2004; Westlund Reference Westlund2009).

Constitutively relational views of practical autonomy are best interpreted as views of autonomous agency rather than autonomous choice or action (Holroyd Reference Holroyd2009, 325–326). An account of autonomous agency specifies the conditions that make autonomous action and choice possible. According to the relationalist, at least some of the necessary conditions for autonomous agency refer to social relations. They may further argue that certain social conditions are both necessary and sufficient for autonomy, but this is typically not taken to be a requirement for a relational view.Footnote ⁴ The account I offer, for example, will incorporate both procedural and relational conditions. Contemporary philosophers who offer constitutively relational views of practical autonomy typically develop them in response to problems with purely procedural views. The general strategy is to construct cases where an individual satisfies the procedural conditions for autonomy, but intuitively lacks practical autonomy due to oppressive social conditions. For example, the deeply deferential wife may reflectively endorse a subservient form of life and therefore meet various procedural conditions, but intuitively she is not autonomous because of her deferential character (Westlund Reference Westlund2009). The same goes for the gaslighted woman (Benson Reference Benson1994). She may have the capacities required for self-reflective evaluation and adjustment of her desires or values, but her lack of trust in those powers undermines her autonomy. While these and other cases discussed in the literature are not without controversy, most relational theorists agree that there are at least some cases where proceduralism fails to account for the ways oppressive conditions can interfere with individual autonomy.

It’s worth noting before moving on that, strictly speaking, only substantive or content-involving views of autonomy directly contrast with procedural views. It is logically possible to have a view that is both formal and constitutively relational, and one practical philosopher of autonomy has defended such a view (see, e.g. Westlund Reference Westlund2009). Since I am primarily interested in the social nature of autonomy and since, as Paul Benson points out, relational views also tend to be substantive views, the distinction is not pressing (Benson Reference Benson and Oshana2015). I will ignore this detail for now and comment on it directly in my concluding remarks.

2. Proceduralism in the intellectual realm

In order to motivate a relational account, it will help to see how procedural views of intellectual autonomy fail in similar ways to practical views. Consider Linda Zagzebski’s procedural account (Zagzebski Reference Zagzebski2013, Reference Zagzebski2015). Although Zagzebski’s goal is to present a more social conception of epistemic agency, the conditions she appeals to are formal in nature. On her view, in order to count as an autonomous agent, one must be capable of self-government. To be self-governed is to play a role in resolving the cognitive dissonance that arises when beliefs come into conflict. To borrow Zagzebski’s example, suppose that before going to bed you believe that you turned off the watering system (Zagzebski Reference Zagzebski2013, 249). When you hear the sprinkler system start up just as you are drifting off to sleep, you are disposed to form a conflicting belief that the watering system is on. This leads to a short-lived state of cognitive dissonance. In order to resolve it, you to give up one of the beliefs (in this case probably the former). While Zagzebski recognizes that we often resolve such conflicts automatically and unconsciously, she puts special emphasis on self-conscious processes:

Procedural accounts of practical autonomy emphasize critical reflection on one’s values or desires. In the same way, Zagzebski’s view of intellectual autonomy emphasizes critical reflection on one’s beliefs. Her view is procedural in that all that is required for autonomous epistemic agency are capacities for self-conscious conflict resolution. In her words, ‘the ability of the self to command itself is just a special case of the more general capacity of a self to reflect upon itself and to make adjustments’ (Zagzebski Reference Zagzebski2013, 257).

Zagzebski nevertheless recognizes some causally relational conditions on intellectual autonomy. She emphasizes how social conditions can result in a lack of trust in one’s self-reflective capacities, which can in turn impair one’s autonomy (Zagzebski Reference Zagzebski2015, 236). However, she doesn’t defend these as placing relational conditions on being an autonomous agent. To be an autonomous agent, on her view, is to have the capacities for critical self-reflection. A lack of self-trust may result in the inability to deploy these capacities, but self-trust doesn’t form part of the constitutive conditions for autonomous agency.Footnote ⁵

3. Intellectual autonomy and making a contribution

The problem with Zagzebski’s view is that one can have capacities for conflict resolution but nevertheless lack intellectual autonomy due to a lack of epistemic confidence. To appreciate the relevance of epistemic confidence, consider how intellectual autonomy requires not only relying on one’s present capacities or modes of thought but also occasionally developing brand-new ones. The latter is central to John Dewey’s own relational account of autonomy and education, and it will help to frame the discussion around his central insight. While reflecting on educational practices as a means of developing individual autonomy, Dewey rejects the apparent tension between what he calls ‘freedom and social control’ (Dewey Reference Dewey1916, 352). He writes:

Rather than equate freedom with self-reliance, Dewey associates freedom with conditions that make certain sorts of social interactions possible. Specifically, he focuses on conditions which allow the agent to contribute to social discourse while at the same time preserving their autonomy. In the context of his educational theory, Dewey takes these conditions to be identical to those that make being a discoverer possible. He argues that the role of the teacher is not to drill in behavior that meets predesigned, fixed standards but to guide the student’s exploration of a given subject matter. The teacher should facilitate or create an environment where a student can approach and solve problems suitable to their present knowledge and capabilities through reflection and experimentation. By doing so, they help develop their student’s autonomy.

Although I don’t defend Dewey to the letter, I take up his central insight. I argue that intellectual autonomy requires being in a position to make one’s own contribution to social epistemic practices. I argue that someone who makes up their own mind has the confidence to rely on their own perspective and develop their own ways of thinking. This doesn’t mean that they never doubt themselves or that they never rely on the contributions of others. It means they take themselves to have the same basic standing and ability as others to make their own. As I argue in sections 4–8, acknowledging this point also allows us to account for the way epistemic injustice undermines autonomy.

4. Conceptual know how

If to be autonomous in a practical sense is to be in a position to determine one’s own will, then to be autonomous in an intellectual sense is to be in a position to determine one’s own beliefs. I argue that the latter requires not only the capacities for resolving epistemic conflicts but also a sense of one’s own standing and ability to rely on those capacities as much as anyone else. The view I defend therefore incorporates both procedural and relational elements. Like those who argue for relational conditions on practical autonomy, my strategy is to identify cases where someone’s reasoning capacities are intact, but intuitively they lack autonomy. So, before making my case, it will help to have a better sense of what capacities are required to resolve epistemic conflicts.

One natural way to start, although I don’t claim any methodological superiority for this approach, is by considering the capacities required to participate in social epistemic practices. Consider the following: in a laboratory at CERN, scientists blast particles through a bubble chamber and analyze the results. During a seminar on human nature, philosophers debate the evolutionary case for gender differences. Driving past a tree, a backseat argument erupts over whether it was an oak or an elm. At a local concert venue, someone tries to find out where the bathrooms are. This is a small sample of the various ways we participate in epistemic practices from the mundane activities of everyday life to the more specialized activities of the science lab and philosophy seminar. By ‘practices’ I mean a more or less well definable way of doing things. Ways of doing things can be more or less localized (my way, your way, our way or their way), more or less context-specific, more or less elaborate, more or less contested, and may be described at courser or finer levels of detail or generality. When made explicit, practices are defined in terms of the norms governing them. To say that practices in general are norm-governed is to say that one’s performances within the practice are subject to normative assessment. There is a right and wrong way to use your soup spoon, a right and wrong way to brush your teeth, and a right and wrong way to inbound the ball in a basketball game. When you don’t follow the norms, you open yourself up to criticism. Epistemic practices in particular will be norm-governed ways of disclosing features of the world. For example, there are correct and incorrect ways of performing a scientific experiment or engaging in a philosophical argument, and when you fail to follow the norms of the practice, you open yourself up to criticism (compare Haugeland Reference Haugeland and Scheutz2002).

As with any norm-governed activity, participation in epistemic practices requires some know how; that is, reliable and resilient abilities to live up to the norms governing the task (Haugeland Reference Haugeland1998, 322; Elzinga Reference Elzinga2018a; compare, Löwenstein Reference Löwenstein2017).Footnote ⁶ The particularly salient kind of know how or skill for participating in epistemic practices is conceptual know how. We can distinguish two basic kinds of conceptual know how (see Elzinga Reference Elzinga2018b).Footnote ⁷ The first is what can be called inferential skill. This is the skill of moving between assertions or, their inner analogues, judgments. To deploy inferential skills is to take one belief or judgment to give a (pro tanto) reason for undertaking another.Footnote ⁸ Javi infers ‘this tree is an elm’ from ‘this tree has serrated leaves’. In doing so, he takes his commitment to the latter as a reason for undertaking a commitment to the former. Someone who knows how to make inferential moves has a reliable and resilient ability to live up to the norms governing the task. This means they have a grasp on what conditions make an inference good and what conditions undermine an inference and that they are able to apply their inferential skills to novel tasks.

The second kind of conceptual skill might be called perceptual skill (compare Kukla Reference Kukla2006). To know how to recognize that something is the case is to know how to give uptake to aspects of the world or to take aspects of the world as providing reasons for belief or perceptual judgment.Footnote ⁹ For example, someone who knows how to read an x-ray is able to pick up on features of x-rays and form accurate perceptual judgments (see, e.g., Snowden, Davies, and Roling Reference Snowden, Davies and Roling2000). As with any skill, the conditions under which perceptual skills are deployed will be varied. Radiologists are reliably and resiliently able to detect features of the image – distinguish noise from signal, normality from abnormality – and make perceptual judgments about their significance. Some perceptual skills are more mundane. I can reliably and resiliently distinguish recyclable from non-recyclable materials, for example. Nevertheless, this is a genuine skill I possess, one that a child, for example, might not and might acquire only with some effort.

Now consider Zagzebski’s suburban example of epistemic conflict resolution. You believe that the watering system is off. When you hear the sprinklers running, you are inclined to believe that the watering system is on. The resulting conflict is resolved when you give up one of the beliefs. Going through this process requires capacities for deploying concepts, like [watering system], [sprinklers] and so on, in response to features of the world and in response to other conceptually articulated representations. Without these basic conceptual skills, you won’t be able to recognize that the sprinklers are on, that this conflicts with your original belief that the watering system was off or adjust your beliefs accordingly. These skills therefore constitute the basic rational capacities that make conflict resolution possible.

Recognizing different aspects of the world and reasoning about different topics will of course require different conceptual capacities, and different agents will have different conceptual skills. One may be able to access aspects of the world that are unavailable to someone else, and one may be able to argue about some matter the other cannot reason competently about at all. In general, however, the autonomous agent will have a broad suite of conceptual capacities that make engaging in ordinary everyday discursive practices possible. If this account is correct, then the procedural aspect of autonomous epistemic agency is the deployment of conceptual know how or skills for giving uptake of reasons provided by the world (perceptual skills) or other beliefs or judgments (inferential skills). If these procedural conditions were sufficient for autonomy on their own, one’s beliefs would be autonomously formed when they result from the deployment of such skills. In other words, you believe autonomously when you believe on the basis of reasons that you have acquired the skills to discern and appreciate.Footnote ¹⁰ I intend to show, however, that while the procedural conditions are necessary, they are not sufficient for autonomy.

5. Hermeneutical injustice

Merely possessing the conceptual skills that make conflict resolution possible is not sufficient for autonomous epistemic agency. The autonomous agent also must have confidence in her standing and ability to participate in epistemic practices on an equal par with other epistemic agents – confidence not only to deploy conceptual skills but to deploy them in novel ways or to develop new ones. To argue the point, I focus on a historical episode, which I will later tailor to my particular ends. The principle case I have in mind are the conditions that result when individuals belonging to a particular social group are victims of hermeneutical injustice.

Hermeneutical injustice arises when members of a particular social group are systematically and unjustly prohibited or otherwise prevented from participating in sense-making practices. Someone who is hermeneutically marginalized is treated as lacking the standing or ability to create new conceptual skills or produce new perspectives on the world. The result is a widespread inability to make sense of significant aspects of the marginalized group’s experience, which affects both dominantly situated knowers and marginalized knowers alike (but see, Pohlhaus Reference Pohlhaus2012; Dotson Reference Dotson2012, Reference Dotson2014). In Epistemic Injustice, Miranda Fricker offers the following case to illustrate the idea (Fricker Reference Fricker2009, 149–152). Before the U.S. women’s liberation movement in the late 60s and 70s, victims of sexual harassment were unable to make sense of their experience to themselves and others. This is because, despite the pervasiveness of the problem and its occasional identification in other contexts in U.S. history, the conceptual skills required to recognize and make inferences about sexual harassment were not widely shared. In fact, the term ‘sexual harassment’ didn’t exist (or at least wasn’t widely used) until it was coined in a consciousness-raising meeting in the mid-70s (Brownmiller Reference Brownmiller1999, 280–281). When it was subsequently marketed through speak outs, leaflet distribution and eventually coverage in mainstream newspapers and magazines, the term became widespread and people, from federal judges to ordinary folks on the job, began to develop the conceptual skills for making sense of sexual harassment. In other words, people began to develop the perceptual skills which allowed them to (more or less) reliably and resiliently recognize instances of sexual harassment and the inferential skills which allowed them to connect the phenomena to job discrimination and broader patterns of oppression. Someone who possesses such skills is in a position to take up a perspective on the world not available to someone lacking such skills, so we might say that women and eventually the broader culture began to develop a new epistemic perspective.Footnote ¹¹

Fricker doesn’t offer a general account of how hermeneutical marginalization works or how women were marginalized in this specific case, but a plausible account is not hard to imagine. A combination of isolation from other women with similar experiences (notice that this is the main problem that consciousness-raising groups are designed to address), the prevalence of contradictory narratives (e.g. the idea that the behavior is merely ‘flirtatious’, a matter of ‘boys just being boys’ or perhaps ‘inevitable’ or otherwise ‘trivial’), the fact that aggressors are typically one’s superiors or at least have the bosses on their side, and other factors, like general norms against women developing and voicing new knowledge which conflicts with the status quo, were sufficient to keep women from developing the resources. According to Fricker:

It could also be said that Brownmiller’s memoir, which recounts the history of second-wave feminism from a firsthand perspective, is a story of women gaining a sense of their own standing and ability to participate as equals in certain epistemic practices (Brownmiller Reference Brownmiller1999). Although Fricker avoids historical speculation about the present case, she argues that one of the potential harms of epistemic injustice is that it undermines one’s epistemic confidence (Fricker Reference Fricker2009, 163; see, also, Roessler Reference Roessler and Oshana2015). In his own work on epistemic injustice, José Medina remarks that extreme forms of oppression can lead to an ‘utter lack of self-confidence and epistemic self-trust’ (Medina Reference Medina2013, 41). In the present case, the idea would be that persistent and systematic exclusion of women from sense-making practices surrounding sexual harassment and from sense-making practices more generally could undermine one’s confidence to develop new conceptual skills when confronted with an experience that has no significance in the broader culture.

Although Fricker does not offer an analysis of the concept, I take it that one can lack intellectual confidence in one of two ways: (1) by lacking confidence in one’s standing to develop new conceptual skills or (2) by lacking confidence in one’s ability to develop new conceptual skills. To lack a sense of one’s standing would be to take oneself to lack the authority to develop one’s own perspective on the world by developing new conceptual skills. Since developing new conceptual skills, on the interpretation of know how and skill adopted here, involves developing new conceptual norms, we can also say it is to take oneself to be unauthorized to make a normative claim – a claim about which conceptual skills are worth having and which epistemic perspectives track objective features of the world. Nevertheless, marginalization may only be ‘more and less persistent and/or wide-ranging’ and ‘someone might be hermeneutically marginalized only fleetingly, and/or only in respect of a highly localized patch of their social experience’ (Fricker Reference Fricker2009, 153). To lack a sense of one’s ability to develop new conceptual skills is to take oneself to be incapable of developing new conceptual skills. Someone might feel incapable of engaging in epistemic projects because they lack the opportunity, because such a project would meet with resistance from others, or perhaps because they doubt their intelligence or resourcefulness. Again, one’s felt lack of confidence, in this case in one’s abilities, may be limited only to some domains and not others.Footnote ¹²

In general, the effects of hermeneutical marginalization will be varied. Neither Fricker nor I wish to deny that there were pockets of resistance or critical consciousness before the women’s liberation movement. Nevertheless, it was because women came together and restored each other’s epistemic confidence that various epistemic injustices were, to some extent, overcome. I submit these claims as hypotheses, but they are plausible and fit the dominant narrative about the history of second-wave feminism. It is worth noting however that there have been a number of constructive criticisms following Fricker’s original work (see Dotson Reference Dotson2012, Reference Dotson2014; Pohlhaus Reference Pohlhaus2012; Medina Reference Medina2013; Berenstain Reference Berenstain2016). Medina, for example, argues for a more pluralistic treatment of epistemic injustice, which emphasizes that a ‘complex social body always contains heterogeneous hermeneutical publics with diverse resources’ (Reference Medina2013, 104). It may therefore be misleading to speak of ‘women’, ‘women’s experience’, and ‘hermeneutical resources’ in monolithic terms. Commenting specifically on the case at hand, Nora Berenstain points out that Fricker fails to acknowledge the significant contributions of black women in facing and addressing the problem of sexual harassment (Berenstain Reference Berenstain2016, fn. 20). These and other valuable critiques aside, the undisputed fact that epistemic injustice has the potential to undermine one’s epistemic confidence is enough to illustrate my point about autonomy, which I turn to now.

6. Fictional case study

Some of Fricker’s own comments indicate how marginalization might inhibit perspective development while leaving one’s rational capacities intact. Fricker routinely describes the experience of a hermeneutical gap in terms of a sense of ‘dissonance’ between one’s own experience and the collectively shared hermeneutical resources (Fricker Reference Fricker2009, 163–168).Footnote ¹³ Recall that Zagzebski’s describes the capacities for autonomous agency in terms of self-conscious resolution of cognitive dissonance. This terminological overlap may be a coincidence, but it’s a suggestive one. Someone who is hermeneutically marginalized may have the conceptual capacities required to recognize and resolve the dissonance that occurs when their experience conflicts with the dominant hermeneutical resources. However, if they have an impaired sense of their epistemic self-worth, they will be disposed to side with the dominant understanding rather than pursue new conceptual resources.Footnote ¹⁴

Recall however that there are two kinds of relational account: those that place causal conditions on autonomy and those that place constitutive conditions on autonomy. To develop a causally relational view, it is enough to describe the ways in which social conditions can prevent the expression or development of individual autonomy. I argue for a stronger thesis. Specifically, while epistemic confidence non-relationally understood is causally useful for developing intellectual autonomy, epistemic confidence relationally understood is constitutively necessary. To clarify why we must add relational conditions to the procedural view outlined in section 3, it will help to contrast a cast of fictional characters based on the current example. The first two characters both exercise their conceptual capacities and develop new ways of thinking. While one develops new ways of thinking because she has epistemic confidence, the other develops new ways of thinking only because she defers to others in her social circle.

Integral to the development of the concept of sexual harassment was the occurrence of various consciousness-raising meetings where women came together to share their experiences and undertake the project of conceptual skill development as a joint project. Suppose that two characters, call them Patty and Claudia, attend such a meeting. Further, suppose that they both possess roughly the same conceptual capacities and come from more or less the same socio-economic background. While they have both experienced some form of sexual harassment in the past, neither has been able to make sense of the experience in explicit terms. When another attendee describes her past experience with sexual harassment, both relate to her story and come to feel a sense of cognitive dissonance between their implicit understanding their experience and the conceptual resources made available by the broader society.Footnote ¹⁵ Through grouping their similar experiences under the term ‘sexual harassment’ and relating the concept to other ideas, both Patty and Claudia begin to develop a new set of conceptual skills. Suppose that Patty develops these conceptual skills because she has confidence in her ability to recognize the conceptual lacuna. She relies on her own sense of cognitive dissonance and the way it drives her to articulate a new perspective on the phenomena. That’s not to say that she doesn’t also rely on the experiences and ideas of the other women in the group. It’s to say that her own experiences and sense-making faculties play a significant causal role in developing the skills and that in favorable circumstances she would be able to develop the relevant skills on her own (at least in some rough and ready form). We may further suppose that this tendency to rely on her own experiences and conceptual skills is a more or less domain-general trait of Patty’s. She does of course rely on other people’s points of view in forming her patterns of thought, but she typically relies on her own experiences too when she can.

Claudia, on the other hand, suffers from severely deficient epistemic confidence, and she develops the conceptual skills for making sense of sexual harassment only because she defers to the other women in the group. This means that she doesn’t rely on her sense of cognitive dissonance, try to resolve it on her own or make any contribution to the group. She relies solely on the testimony of the other women and how they go on to make sense of the experience. We may suppose that she nevertheless genuinely gains the conceptual skills along with them because she is generally highly motivated to conform to the (epistemic) behavior of those around her. Finally, if she was instead in a group dominated by men or women inclined to say that experience of the woman who spoke up was merely a matter of ‘boys being boys’ or something of the sort, she would have gone along with that analysis despite her contrary experiences. We may further suppose, to strengthen the case, that Claudia’s deferential attitude is a persistent, domain-general trait and that even when she is on her own, she tries to believe as she thinks others would have her believe in that situation. This means that even when she forms her beliefs without anyone to directly defer to and, moreover, even if her deferential attitude leads to true or justified beliefs, she does not make up her own mind. What I want to suggest, with some qualifications to come in the remaining sections, is that Patty participates in epistemic practices with intellectual autonomy and Claudia does not. If this is so, then epistemic confidence is constitutively necessary for intellectual autonomy. Lack of confidence doesn’t just prevent Claudia from deploying the capacities that make believing autonomously possible. Even if she does deploy her conceptual capacities and develop new skills, if she does so without epistemic confidence, she does not believe autonomously.

7. Epistemic confidence and deep deference

This last point requires further elaboration. Someone who lacks epistemic confidence in their conceptual capacities is deferential as a result, and their deferential attitude apparently undermines their intellectual autonomy. Consider an epistemic analogue of what Andrea Westlund calls the deeply deferential agent. Westlund characterizes such agents as follows:

Westlund is interested in agents who are deferential about their reasons for choosing or acting as they do, and she argues that they lack practical autonomy because they are ‘in the grip’ of a reason-giving policy that is not their own. In response, Westlund claims that we must acknowledge a constitutively relational condition on practical autonomy according to which the autonomous agent ‘holds herself answerable, for her action-guiding commitments, to external critical perspectives’ (Westlund Reference Westlund2009, 35).

Whether or not this move is successful in the practical case needn’t overly concern us here.Footnote ¹⁶ The immediate problem for making this move in the intellectual case is this. Given recent trends in social epistemology, it may not even be obvious why deference should be incompatible with intellectual autonomy in the first place. Testimonial knowledge is, after all, deferential knowledge – it characteristically involves, as Sanford Goldberg puts it, ‘pass[ing] the epistemic buck’ – and most agree that we can autonomously rely on the word of others (Goldberg Reference Goldberg, Lackey and Sosa2006, 134). Moreover, rejecting epistemic deference outright would land the relational theorist right back where we started with an individualistic view of intellectual autonomy. This would be quite the ironic step in the wrong direction for someone after a relational theory (especially since epistemic deference is already compatible with autonomy on procedural views like Zagzebski’s).

Because deference isn’t necessarily incompatible with autonomy, it’s important to emphasize that our concern is with deep deference. A deeply deferential agent is not just deferential but also deferential about their deference. A deeply deferential epistemic agent would be someone who, first, nearly always defers to someone else when prompted back up their beliefs or their way of acquiring them and, second, if they are pressed to defend their deferential behavior, they point to someone else’s reasons rather than providing independent grounds of their own. The latter behavior isn’t always autonomy undermining. I might defer to an expert about some matter and when prompted to defend my deference appeal to other experts or meta-experts who vouch for their expertise (see, e.g., Goldman Reference Goldman2001). It’s only more or less persistent and widespread deference that undermines intellectual autonomy. In short, ordinary forms of deference may not be incompatible with autonomy, but deeply deferential behavior is clearly out of step with the idea of intellectual self-governance.

Claudia lacks epistemic confidence in the sense that she is deeply deferential. As a result, she may reflectively underestimate her intellectual abilities, but this doesn’t capture the relevant sense in which Claudia lacks confidence. It will help to distinguish two general senses of confidence. Suppose that someone talks a big game about their abilities, but when it comes to actually performing they under no circumstances perform with confidence. They are hesitant, slow to react, try to keep out of sight and so on. We can say that they have high evaluative confidence, i.e., they sincerely but mistakenly evaluate their abilities highly, but they lack performance confidence, i.e., they are not confident performers.Footnote ¹⁷ The above reflections on deep deference indicate that epistemic confidence in the relevant sense is a type of performance confidence. Someone who is epistemically confident is disposed to participate in epistemic practices with confidence. Someone who lacks epistemic confidence does not, and as a result they are deferential. The more they lack confidence the more deferential they will be, and so we might say that to be deeply deferential is to be deeply epistemically diffident.Footnote ¹⁸ The claim that Claudia lacks epistemic performance confidence in making sense of sexual harassment is nevertheless consistent with the claim that she genuinely possesses the skills for making sense of sexual harassment. To have a skill is to have a reliable and resilient ability, and one may possess an ability without being robustly disposed to deploy it. Claudia, for example, may only deploy her skills in the presence of certain others or, perhaps when pressed to defend her conceptual practices, she readily defers to other feminists instead of offering reasons of her own even though she could.Footnote ¹⁹

One needn’t be as confident as Patty to have intellectual autonomy or as diffident as Claudia to lack it. It’s also important to recall in this connection that I am after a necessary condition for counting as an autonomous agent rather than the conditions which must be met for a particular belief to count as autonomously formed. The autonomous agent will have a certain character trait of epistemic confidence. They may not exhibit confidence in all circumstances, they may occasionally be epistemically lazy or satisficing, and they will often defer to experts whose opinions they take to be reliable. The important feature of the autonomous agent, however, is that they don’t systematically undervalue their own conceptual skills, ability to recognize epistemic conflicts or dissonance, and develop new skills in response. There may not be any precise metric for deciding how persistent and widespread deferential behavior or lack of performance confidence has to be to be autonomy undermining, but since autonomy is plausibly thought of as something that comes in degrees, this may not be a problem. In any case, I limit myself to the claim that some degree of epistemic confidence is necessary for intellectually autonomous agency, and I leave the possibility of a principled cutoff point open.

8. Causal and constitutive conditions

The present case allows us to see that intellectual autonomy requires having a sense that one has the same basic capacities for recognizing and resolving epistemic dissonance as anyone else and that one has just as much standing as others to rely on one’s conceptual skills in order to develop new ways of thinking. In short, it requires taking oneself to have equal standing and ability to deploy and develop conceptual skills in relation to other epistemic agents. The point about equality, especially that of equal ability, is easily misunderstood. To say that one takes oneself to have equal standing as others is simply to say that you don’t take yourself to be barred from thinking through an issue and relying on your perspective in comparison with others. To say that one has the sense that one is equally capable as others to deploy and develop conceptual capacities is not meant to imply that one takes oneself to have the very same specific conceptual skills as anyone else. It does require, however, having the sense that one is more or less able to participate in practices of deploying and developing conceptual skills. I don’t take myself to be as able to think about particle physics in just the same way as the physicist, but I do have the sense that I have the same basic ability to participate in epistemic practices as her. I have the sense that can think through most ordinary, everyday topics, that I have some specialized knowledge of my own and that, if I tried, I could learn the language of particle physics at least to some extent.

The relational character of the condition I propose falls out of the case of Patty and Claudia discussed above. What undermines Claudia’s autonomy is not that she lacks epistemic confidence per se, but that she participates in epistemic practices while taking herself to be inferior to other epistemic agents. To underwrite this point, consider another character. Catalina, let’s call her, is again like the other characters I’ve described so far in terms of conceptual capacities and personal history, but she is skeptical about the possibility of anyone developing the conceptual resources for making sense of sexual harassment. She lacks performance confidence with respect to the project, but not in the relevant relational sense. Her view is that it simply cannot be done, not that she is less able or permitted to do it than others. As a result, she offers more criticisms than constructive proposals, dwells on the complexity or difficulty of the task and suggests devoting resources elsewhere. It’s not that she has an ulterior motive for her skepticism. Instead, she has a hard time reconciling the currently dominant conceptual resources with what the other women’s experience and imagining new conceptual possibilities. Her situation is rather like that of an athlete burdened with deeply entrenched habits which make innovation hard to imagine or the complete novice who, justifiably or not, uses their unfamiliarity and frustration with a task as evidence that it cannot be done.

Because of her skeptical attitude, Catalina has low confidence in her ability and standing to engage in the project, but she takes herself to be equal to others in this respect. Her lack of confidence is nevertheless potentially autonomy undermining because it may prevent her from relying on her experience and developing concepts that counter the dominant narrative about sexual harassment. To the extent that she fails to develop a perspective from which to resist, challenge or criticize the dominant narrative, her thought will be under the control of the broader societies interpretation. Sometimes it is possible however to lack confidence and accomplish something anyway (for example, in those cases where we feel we have no other option but to try). Suppose that Catalina along with the other women in the group nevertheless come to make sense of the phenomena, and Catalina, spurred on by the rest of them, makes up her mind on the matter too. Specifically, once she sees how the other more optimistic and persistent women draw on their collective experiences to start making sense of the concept, she begins to chime in and at least implicitly recognize that she can draw on her experience in some way (or if she doesn’t actually speak out loud in the group, she at least begins to draw on her experience to make sense of it for herself). I am inclined to say that she is intellectually autonomous even though she lacks confidence (and this is so whether or not she gains confidence through the exercise). Her lack of confidence will make the project hard, but that doesn’t mean she fails to make up her own mind. It was a struggle, but she did it. She pulled through. Moreover, she makes up her mind precisely in the sense that she relies on her own experience to develop the concept and makes her own contribution. Unlike Claudia, she doesn’t develop the conceptual skills only because she defers to others.

Non-relational epistemic confidence, the confidence that you can or may participate in epistemic practices in some way, is therefore a helpful characteristic of the epistemic agent that, when absent, could be supplemented in other ways without undermining autonomy. On the other hand, recall that Claudia was like Catalina in that she was able to understand the nature of sexual harassment, but, unlike her, she rated herself as inferior to other epistemic agents with respect to her epistemic agency. This means that even if she manages to develop new ways of thinking, she doesn’t do so autonomously. This implies that only relational epistemic confidence, the sense that one is equal to other epistemic agents, is constitutively necessary for intellectual autonomy. This does not, however, mean that having a relational sense of confidence isn’t also causally useful. In fact, it is precisely such confidence through which, in the case of sexual harassment, women recognize themselves as having the standing and ability to shape and develop communal resources alongside or in opposition to men in order to mitigate the effects of hermeneutical injustice.

9. Comparisons with practical accounts

Before concluding, it will help to draw parallels with relational accounts of practical autonomy and explore potential objections to both. The present account of intellectual autonomy is not just constitutively relational. It also places what are called substantive conditions on autonomy. In direct contrast to procedural or content-neutral views, substantive or value-laden views place constraints on the contents of an autonomous agent’s values or preferences. On strong substantive views, ‘the contents of the preferences or values that agents can form or act on autonomously are subject to direct normative constraints’ (Benson Reference Benson and Stacey Taylor2005, 133). Marina Oshana, for example, argues that one cannot adopt subservient or slavish preferences and retain one’s autonomy (see, e.g., Oshana Reference Oshana2006, Reference Oshana and Oshana2015; Stoljar Reference Stoljar, Mackenzie and Stoljar2000). Oshana’s view is also externalist in the sense that satisfying conditions external to an individual’s psychology (for example, being embedded in an environment which affords a sufficient range of genuine opportunities) are necessary for autonomy. Weak substantive views also constrain normative content, yet they do not place direct normative constraints on an agent’s attitudes (Benson Reference Benson and Stacey Taylor2005). Paul Benson, for example, argues that practical autonomy requires a sense of one’s self-worth as an agent, and this requirement indirectly constrains the contents of one’s attitudes and the sort of life one can autonomously live.Footnote ²⁰ Benson’s self-worth condition is therefore internalist (as are procedural views) because autonomy relevant conditions only refer to features of an individual’s psychology.

Like weak-substantive, internalist views of practical autonomy, the present account of intellectual autonomy constrains the sort of self-regarding attitudes an autonomous agent can have and thereby indirectly places restrictions on how autonomous agents form their beliefs. As such, it is more restrictive than procedural views, but less restrictive than a strong substantive view of intellectual autonomy would be. To see what a strong substantive view might look like, consider one final case. Suppose that Lillian self-consciously adopts an epistemically subservient form of life. She believes that women shouldn’t take an active role in the production of new knowledge, and so the content of her attitudes is in direct opposition to the value of intellectual autonomy. A view that seeks to accommodate this alternative intuition would perhaps, like the strong substantive view of practical autonomy, place direct restrictions on the way epistemic agents form their beliefs.

Whether or not this is the only way to accommodate the alternative intuition, the comparison with strong and weak substantive views of practical autonomy is instructive. There is an apparent tension that results when we try to develop accounts of autonomy sensitive to oppressive conditions (see Khader Reference Khader2011; Mackenzie Reference Mackenzie and Oshana2015). On the one hand, we want our account to help us make sense of the ways oppressive conditions undermine, impair or damage individual autonomy. On the other hand, we don’t want to run the risk of disrespecting or underestimating the agency of oppressed individuals or ruling out the self-ownership of their choices simply because they are misguided. In the current state of the literature on practical autonomy, strong substantive theorists are accused of favoring the former goal over the latter, and weak substantive theorists are accused of doing the reverse (Warriner Reference Warriner and Oshana2015, 25–32). Like weaker views, my account of intellectual autonomy puts emphasis on respecting the oppressed individual’s epistemic agency. So long as Lillian’s acceptance of an epistemically subservient life doesn’t result in deep deference, she still counts as autonomous on my view. In other words, if she is prepared to give reasons in favor of that lifestyle and respond to potential criticisms without deferring, she still retains intellectual autonomy. As such, I leave the possibility of choosing subservience with intellectual autonomy open, just as weak practical views leave the possibility of autonomously choosing a subservient form of life open.

One may further expand this line of thought to respond to extant criticisms of weak-substantive views as they might apply to the present account. Jennifer Warriner, for example, argues that someone may satisfy Benson’s conditions while following an oppressive gender-script provided by the dominant culture, which thereby intuitively undermines their autonomy (Warriner Reference Warriner and Oshana2015, 33–39). Applied to the present case, imagine someone who satisfies my criteria and follows an ‘epistemic script’, roughly a set of norms governing the deployment of a concept, handed down from an oppressive community. One may argue that this is indeed imaginable. Take, for example, Patty before she has joined the feminist movement and seriously considered the issue of sexual harassment. Suppose that she is epistemically confident but that she nevertheless follows oppressive scripts which lead her to see instances of sexual harassment in the wrong way (e.g. as a matter of ‘boys being boys’). One might think that my view misclassifies Patty as intellectually autonomous.

Although my view does imply that Patty is autonomous, this is the correct result. According to the view defended here, to be intellectually autonomous is to be prepared to revise communal epistemic scripts in light of epistemic conflicts. Even if Patty’s views are misguided, if she is confident in her ability and standing to revise her perspective in the appropriate circumstances, she is intellectually autonomous. She follows an oppressive script, but she is not under the control of that script in the relevant sense for intellectual autonomy. Someone who lacks autonomy, on the other hand, like Claudia, is genuinely under control of the culturally dominant scripts because she will not rely on her own sense of cognitive dissonance to revise them when the time comes. Therefore, while my view doesn’t require having the right epistemic perspective, it does require being in a position to improve one’s epistemic perspective.

10. Conclusion

Paul Benson and Diana Meyers have recently argued that the debate between strong and weak views of practical autonomy can only be settled by taking a closer look at the role the concept plays, or should play, in our real-life normative projects (see Benson Reference Benson, Veltman and Piper2014; Meyers Reference Meyers, Veltman and Piper2014). My goal in this paper has been to broach the topic of relational intellectual autonomy by putting forward a concrete proposal rather than providing a survey of the current literature and exploring potential avenues for further research. I hope this can open up further conversations about relational views of intellectual autonomy, as well as the importance of the concept and its role in our social and political projects. With that said, there are a couple of quick reasons to be optimistic about the present account. For one, it respects the possibility of intellectual autonomy in the face of oppressive social influence by allowing for subservient attitudes. At the same time, it leaves room for critical engagement with the social conditions that give rise to deep deference, as long as deep deference inhibits the production of knowledge. As Benson points out, there is a difference between autonomy (self-rule) and orthonomy (right-rule). Oppressive social relations that don’t inhibit the former can still be criticized for inhibiting the latter (Benson Reference Benson and Stacey Taylor2005, Reference Benson, Veltman and Piper2014).